The oldest old are one of the fastest growing segments of the population, but we know little about the risk factors for mortality at the oldest ages. Only a few studies have examined risk factors for those over age of 85, and none have examined whether those factors increase or decrease in importance in the age range 85-100. Models of mortality in heterogeneous populations have shown how variation in mortality among individuals can lead to misleading conclusions about declines in the importance of individual risk factors. The proposed research will examine this issue using a data set that combines information from three surveys: the National Longitudinal Mortality, the Health and Retirement Study, and the AHEAD study. Combining the observations for persons aged 60 and over in these surveys will provide a sufficiently large sample to examine in detail how the relative risks of fixed traits change with age. The combined data set also covers a sufficiently long period that it will be possible to examine changes in risk ratios within cohorts over a period of almost 20 years. The aims of the research are to 1) combine the surveys into a single data set, 2) study changes over age in relative risks for fixed traits in periods and cohorts, 3) compare changes in relative risks with models based on fixed frailty, and 4) test models based on two or more risk factors. The models will be used to study changes in the effects of race/ethnicity, education, marital status (never married), and place of birth.